clopen.solutions
OP-00702 Solution claimed

Removing the logarithmic loss in Helmholtz Nyström estimates

Submit a result

Let $\Gamma \subset \mathbb{R}^2$ be a smooth embedded closed curve with a smooth periodic parametrisation $\gamma : \mathbb{R}/2\pi\mathbb{Z} \to \Gamma$ satisfying $0 < |\gamma'(t)| < c_{\max}$. For $k > 0$ consider the oscillatory kernels $$L_{1,k}(t,\tau) = k \int_{S^1} e^{ik\langle \gamma(t) - \gamma(\tau), \omega\rangle} f_k(\omega, t, \tau)\, dS(\omega),$$ where the amplitude $f_k$ is smooth, periodic in $t$ and $\tau$, and has every derivative in $(\omega, t, \tau)$ bounded independently of $k$; kernels of this form arise from the Helmholtz single- and double-layer operators in the splitting behind Kress quadrature. Let $$\widehat L_{1,k,m}(t) = \frac{1}{\sqrt{2\pi}} \int_0^{2\pi} e^{-im\tau} L_{1,k}(t,\tau)\, d\tau, \qquad \langle z \rangle = (1 + |z|^2)^{1/2},$$ regard functions of $t$ as functions on $\Gamma$, and let $H^s_k(\Gamma)$ carry the norm $\|u\|_{H^s_k} = \|(I - k^{-2}\Delta_\Gamma)^{s/2} u\|_{L^2(\Gamma)}$. Galkowski, Rachh and Spence proved the bound below when $\Gamma$ is convex with nonvanishing curvature and $\gamma$ has unit speed. For general $\Gamma$ they proved it with $C$ replaced by $C\sqrt{\log k}$, for each fixed $\epsilon$ and $N > (1+\epsilon)(1-\epsilon)^{-1} c_{\max} k$.

Problem. Is it true that for every $k_0 > 0$ and $s \in \mathbb{R}$ there is a constant $C$ such that $$\frac1k \sum_{0 < |m| \le (1-\epsilon)N} \|\widehat L_{1,k,m}\|_{H^s_k(\Gamma)} \langle m/k \rangle^s \le C$$ for all $k > k_0$, $N > 0$ and $0 < \epsilon < 1$? The constant may depend on the curve, the parametrisation, the bounds on the amplitude, $s$ and $k_0$, but not on $k$, $N$ or $\epsilon$.

Source: Galkowski, Rachh and Spence (2025), Conjecture 11.4. Stated by AIM — Open Applied Problems for an arbitrary smooth amplitude, with the quantity of the conjecture written out.

Imported from github.com/MColbrook/AIM.

Classification

Primary
NA — Numerical Analysis
Secondary
AP
Keywords
AIM Open Applied Problems, boundary integral equations, Helmholtz equation, Nyström method

References

@misc{galkowski2025,
  author        = {Galkowski, Jeffrey and Rachh, Manas and Spence, Euan A.},
  title         = {Helmholtz boundary integral methods and the pollution effect},
  year          = {2025},
  eprint        = {2507.22797},
  archivePrefix = {arXiv},
  note          = {Version 3, 21 March 2026},
}
@misc{aim-nystrom-logarithmic-loss,
  author       = {Brunton, Steve and Colbrook, Matthew J. and de Hoop, Maarten V. and Stepaniants, George and Townsend, Alex and Ward, Rachel},
  title        = {{Removing the logarithmic loss in Helmholtz Nyström estimates}},
  howpublished = {AIM --- Open Applied Problems, GitHub repository},
  year         = {2026},
  url          = {https://github.com/MColbrook/AIM/blob/1b7c7fd251c1643bb41f0e3cef38704bcef410fd/problems/523-nystrom-logarithmic-loss.md},
  note         = {Entry 523 at commit 1b7c7fd, 2026-10-13},
}

Resolution claims

Siavash Sadeghi

A fixed smooth amplitude on a smooth closed curve with a straight arc gives growth of order square root of log k. This addresses the arbitrary smooth amplitudes and general parametrizations allowed in the target; it does not assert a counterexample for every physical boundary-integral kernel.

Counterexample · not verified · fidelity not established · posted on GitHub 4 Oct 2026

Status history

WhenChangeByReason
2026-10-16 Needs verification → Solution claimed — Recorded here: AIM @ a153f46, siavash-sadeghi-523/aim523_report.pdf.

Citing this problem

This identifier is permanent. It will not change if the problem is solved, reclassified or withdrawn.

OP-00702, "Removing the logarithmic loss in Helmholtz Nyström estimates", clopen.solutions.